Metabolic phenotype mediates the outcome of competitive interactions in a response‐surface field experiment

Abstract Competition and metabolism should be linked. Intraspecific variation in metabolic rates and, hence, resource demands covary with competitive ability. The effects of metabolism on conspecific interactions, however, have mostly been studied under laboratory conditions. We used a trait‐specific response‐surface design to test for the effects of metabolism on pairwise interactions of the marine colonial invertebrate, Bugula neritina in the field. Specifically, we compared the performance (survival, growth, and reproduction) of focal individuals, both in the presence and absence of a neighbor colony, both of which had their metabolic phenotype characterized. Survival of focal colonies depended on the metabolic phenotype of the neighboring individual, and on the combination of both the focal and neighbor colony metabolic phenotypes that were present. Surprisingly, we found pervasive effects of neighbor metabolic phenotypes on focal colony growth and reproduction, although the sign and strength of these effects showed strong microenvironmental variability. Overall, we find that the metabolic phenotype changes the strength of competitive interactions, but these effects are highly contingent on local conditions. We suggest future studies explore how variation in metabolic rate affects organisms beyond the focal organism alone, particularly under field conditions.


| INTRODUC TI ON
An individual's metabolic rate and competitive ability are tightly linked (Biro & Stamps, 2010;Sloman & Armstrong, 2002). Competition for resources among and within species constrains the acquisition of energy by individuals and, consequently, their growth and fitness.
Similarly, an individual's metabolic rate is tightly linked to its resource demands, with individuals with higher metabolic rates having greater resource requirements (Brown et al., 2004;Burton et al., 2011).
At high population densities especially, competition limits per capita resource availability (exploitative competition) or access to resources (interference competition) (Antonovics & Levin, 1980;Violle et al., 2010). Accordingly, individuals sometimes downregulate their metabolic rates in response to intraspecific competition so as to maintain positive energy fluxes despite lower resource availability (DeLong et al., 2014;Ghedini et al., 2017; but see Yashchenko et al., 2016). Yet competitive environments do not always favor lower, "more thrifty" metabolic phenotypes. under natural, more variable conditions are a necessary next step (Álvarez & Nicieza, 2005).
The competitive advantages conferred by any one metabolic phenotype should depend strongly on environmental conditions (Killen et al., 2013). A higher metabolic rate is often associated with a higher energy turnover that can be beneficial for growth and survival when resource availability is high but disadvantageous when resource levels are low Auer et al., 2015Auer et al., , 2020Bochdansky et al., 2005;Burton et al., 2011;Killen et al., 2011). Individual differences in metabolic rate and associated traits such as body size may therefore interact with environmental conditions to determine the outcome of competitive interactions among conspecifics. For example, individuals with higher metabolic rates may only be able to grow larger and, therefore, be competitively dominant if the per capita resource levels are high (Ward et al., 2006). Conversely, higher metabolic rates may be disadvantageous in resource-limited environments if resources are simply insufficient to sustain individuals with higher energy demands (Auer et al., 2020). How environmental conditions interact with metabolic rate to determine competitive outcomes in the field, however, remains poorly understood.
In a manipulative field experiment, we examined how metabolic rate mediated conspecific interactions in the bryozoan, Bugula neritina, a colonial, sessile marine invertebrate. Colonies of B. neritina are commonly found as part of fouling communities throughout the world where they form dense congregations of conspecifics, which may result in intense competitive interactions among individuals.
We took advantage of the natural and persistent variation in metabolic rate among individual B. neritina colonies (Pettersen et al., 2016(Pettersen et al., , 2020Schuster et al., 2019Schuster et al., , 2021 to test for the effects of metabolic rate on competition. We used a trait-specific responsesurface design to create pairwise interactions of individuals with differing metabolic rates Inouye, 2001). Due to the sessile nature of B. neritina, we were able to follow the performance of individuals across their entire lives in the field. We then measured the outcomes of pairwise interactions by comparing the performance (survival, growth, and lifetime reproductive output) of individuals with different metabolic rates, both in the presence and absence of a neighbor colony.
Each ovicell broods a single larva, which is released into the plankton once embryogenesis is complete. Upon release, the nonfeeding larvae are immediately competent to settle, and most larvae settle within hours under field conditions (Burgess & Marshall, 2011).
Larvae also preferentially settle close to conspecifics in the laboratory (Keough, 1984), and such aggregations of B. neritina conspecifics are often observed in the field.
To obtain individuals for our experiments, we spawned colonies according to standard procedures (Schuster et al., 2019). Briefly, we kept colonies in the laboratory in field-collected seawater in aerated tanks in the dark. After 48 h, we spawned colonies by exposing them to bright light and settled single larvae in a drop of seawater on roughened A4 acetate sheets to induce settlement (~150 settlers per acetate sheet). After 3 h, we rinsed unsettled larvae from the acetate sheets and kept settlers in tanks with unfiltered seawater.
The next day, we attached the A4 acetate sheets bearing settlers to five PVC backing panels (57 × 57 × 0.6 cm; two acetate sheets per panel) and suspended the panels 1 m below the water surface with settlers facing down at the Royal Brighton Yacht Club (37°54′25″S, 144°58′54″E).

| Mass-independent metabolic rate
To conduct metabolic rate measurements, we returned acetate sheets bearing settlers to the laboratory after they had been in the field for 2 weeks. We kept colonies in aerated tanks with fieldcollected seawater at 19°C overnight. Prior to metabolic rate measurements, we removed any epibionts and debris from the colonies.
We then separated individual colonies from the A4 sheets by cutting around the base of the colonies such that each colony was attached to a small square of acetate sheet. In total, we measured the metabolic rates of 372 colonies using 750µl glass vials (Loligo Systems, Denmark) and 24-channel PreSens sensor dish readers (SDR2, PreSens, Germany). We determined metabolic rates as oxygen consumption rates at 19°C over 3 h as described in Schuster et al. (2019). We then converted estimates of V̇O 2 (ml h −1 ; White et al., 2011) to metabolic rates (mJ h −1 ) using the calorific conversion factor of 20.08 J ml −1 O 2 (Crisp, 1971).
We estimated size-independent metabolic rates (MI-MR) by regressing metabolic rate on colony size (nonlinear regression of the form MR = a * M b , where MR is metabolic rate, M is colony size, a is the intercept, and b is the scaling exponent) and extracting the residuals. To determine colony size, we counted the number of zooids in each colony. Given that colonies were attached to squares of acetate sheet, zooid counts were more reliable than weighing them, and the number of zooids and colony mass are strongly correlated (Schuster et al., 2019). Colonies used for metabolic rate measurements ranged from 16 to 48 zooids in size.

| Experimental design and field deployment
Our main goal was to investigate whether metabolic rate mediates the outcome of pairwise interactions using a trait-specific, responsesurface design Inouye, 2001). Based on the continuous range of metabolic rates from our source population (1.25-7.67 mJ h −1 absolute metabolic rates), we generated pairwise combinations of metabolic rates (maximum difference in MI-MR between pairs: 5.19; maximum absolute differences in metabolism between pairs: 5.08 mJ h −1 ; Figure 1). To create our treatments, we glued two acetate sheet squares, each bearing a single colony, onto PVC plates (55 × 55 × 3 mm) such that colonies grew at a distance of 1 cm from each other. We treated both these colonies as the focal colony and neighbor colony to test for reciprocal interactions (Inouye, 2001). In addition, we estimated the baseline relationship between metabolic rate and performance of single colonies without a neighbor colony by gluing a blank acetate sheet square 1 cm from a focal colony (distance between the center of the empty acetate square and the focal colony). We then distributed a total of 162 plates across the five backing panels and redeployed them into the field. Within each panel, we attached plates at a distance of at least 5 cm from each other in order to minimize competitive interactions with colonies on neighboring plates. It is noteworthy that we assigned colonies haphazardly to each panel. Consequently, there were no differences in focal colony size or MI-MR between panels (mean ± SE; Panel 1: zooids: 29.1 ± 6.9, MI-MR: 0.08 ± 0.23; We followed the performance of 265 colonies of known metabolic rates throughout their entire life history, until all colonies had died (April through to October 2019). We followed the survival, growth, and reproductive output of each colony every 2 weeks.
Colonies were considered alive if they were still attached to the plate and >10% of the colony contained feeding zooids. We measured the reproductive output of each colony by counting the number of ovicells throughout the duration of the experiment, and growth as the number of bifurcations at each measurement point (Keough & Chernoff, 1987). We also removed any nonexperimental settlers (both Bugula and other species) from the plate at each measurement point to eliminate competition from other organisms. Furthermore, to avoid any environmental effects and effects from neighboring colonies on other plates associated with a focal colony's position within a panel (i.e., in the center surrounded by lots of competitors vs. at the edge) on metabolic rates and performance, we moved each plate to a different position within the assigned panel every 2 weeks (Mitchell-Olds & Shaw, 1987;Rausher, 1992).

| Statistical analyses
We conducted two different sets of analyses using generalized linear models (GLMs) and repeated measures analyses of covariance (RM ANCOVA): we tested (i) the effects of neighbor colony presence (denoted "1") or absence (i.e., colonies grown in isolation; denoted "0"), and (ii) the effects of neighbor MI-MR and focal MI-MR on focal colony survival at 20 weeks (c.f. Pettersen et al., 2016Pettersen et al., , 2020, growth, the per capita reproductive output over time, and the cumulative reproductive output after 24 weeks (i.e., an individual's summed reproductive output across the life history), respectively. For survival, we conducted a binomial GLM with a logit-function, with focal MI-MR F I G U R E 1 Schematic of the trait-specific, response-surface design used to test the effects of metabolism on pairwise interactions between Bugula neritina colonies. The orange points show the combinations of focal and neighbor colony massindependent metabolic rates (MI-MR) used in pairwise interactions (n = 206); the green points show the mass-independent metabolic rates of colonies grown without a neighbor colony (n = 59). The gray-dashed line indicates equivalences between focal and neighbor MI-MRs colony presence/absence (categorical fixed effect; denoted "1" or "0") or neighbor MI-MR (continuous fixed effect) included in the model. For growth analyses, we also included initial colony size as a fixed effect in analyses to account for differences in focal colony sizes at the start of the experiment. As the response variable, we used either size (number of zooids; log 10 -transformed prior to analyses) or the per capita reproductive output (log 10 -transformed prior to analyses) of focal colonies at each measurement point, respectively.
To derive colony size in terms of number of zooids within a colony (for colonies >2 weeks of age), we converted the number of bifurcations to zooid number by assuming an average of four pairs of zooids in between branching points of a colony (Keough & Chernoff, 1987

| The effect of neighbor colony presence/ absence on focal colony performance
After 20 weeks in the field, we found that focal colony survival  On two panels, colonies grown in the absence of a neighbor colony produced on average 46% more offspring than colonies that were grown in the presence of a neighbor (Figure 4b). Focal MI-MR did not affect cumulative reproductive outputs of focal colonies (F 1,254 = 2.54, p = .11), although focal colonies with lower metabolic rates tended to have higher reproductive outputs.

(a) (b)
Our results pertaining to the effects of neighbor colony presence/absence are summarized in Tables 1 and 2.

| The effect of neighbor MI-MR on focal colony performance
Survival of focal colonies depended on the metabolic rate of both the focal colony and the neighboring colony, and these effects varied among panels (panel × focal MI-MR ×neighbor MI-MR: χ 2 = 5.33, df = 1, p = .02). On one panel, we found that focal colonies with a lower metabolic rate survived better if they were paired with a low metabolic rate neighbor colony (Figure 5a; Panel 4). On the other panels, neighbor metabolic rate affected focal colony survival on some panels but not on others (panel ×neighbor MI-MR: χ 2 = 7.29, df = 1, p = .007). On two panels, we found a positive relationship between neighbor MI-MR and focal colony survival ( Figure 5b; Panels 1 and 2), but on the other two panels there was no effect of neighbor MI-MR (Panels 3 and 5; χ 2 = 0.23, df = 1, p = .63).
Neighbor metabolic rate affected the growth of focal colonies,  TA B L E 1 Summary of survival, growth, and reproductive outputs, and the various effects of neighbor colony presence/ absence across all experimental panels. The significance levels of neighbor colony presence/absence effects within each panel are presented in Table 2 F I G U R E 5 The effect of (a) focal MI-MR and neighbor MI-MR and (b) neighbor MI-MR on focal colony survival probability on panels 1, 2, and 4. Black dots show underlying data points. (a) Warmer colors depict a higher predicted survival probability. Lines in (b) show predicted survival probability from generalized linear models (±95% CI) (Panels 1 and 4) but dissipated on two other panels (Panels 3 and 5).
On one panel (Panel 2), focal colonies paired with a low metabolic rate neighbor colony reproduced more during early stages, but the effect changed in sign at 10 weeks and focal colonies paired with a high metabolic rate neighbor reproduced more thereafter.
Neighbor MI-MR also affected the cumulative reproductive output of focal colonies, but, again, the effect differed across panels Our results pertaining to the effects of neighbor metabolic rate are summarized in Table 3.

| D ISCUSS I ON
We found that the metabolic phenotype of conspecific neighbors altered the performance of focal colonies, but these effects showed strong microenvironmental variability. On most panels,

F I G U R E 6
The effect of neighbor MI-MR on the predicted relative colony size (in zooids; standardized to the mean) over time (in weeks). Black dots show underlying data points. Warmer colors depict higher relative colony sizes Had we not replicated in space, we would have concluded that the effects of neighbor metabolism are more consistent than they actually are. This small spatial variation in the effects of metabolism on performance is likely to maintain variance in this trait (Lange et al., 2016).
We found that individuals with higher metabolic rates grew less than individuals with lower metabolic rates. Recent studies have shown that individuals with lower metabolic rates often grow more and reach larger body sizes due to their relatively lower maintenance costs (Burton et al., 2011;Pettersen et al., 2018), particularly when per capita resource availabilities are scarce (Auer et al., 2015(Auer et al., , 2020Reid et al., 2011Reid et al., , 2012Zeng, Zhang, et al., 2017).
Thus, within a population where conspecifics compete for resources, a lower metabolic rate may confer a growth advantage when resources are limiting. Note: Yellow indicates the response variable increased with neighbor metabolic rate; blue indicates the response variable decreased with neighbor metabolic rate. Green indicates an interaction between both neighbor and focal metabolic rate. Color gradients indicate a change in sign of the effect over time, for example, a change from a negative to a positive effect (blue to yellow) or a change from a negative to no effect (blue to white). We used simple main-effects tests to determine the significance of main effects (and their interactions) within each panel.

Neighbour MI-MR
Colonies growing in the presence of neighbors with lower metabolic rates, which grew to larger colony sizes, tended to grow and reproduce more themselves on some panels. This benefit may have arisen for one of two reasons: (1) neighbors with lower metabolic rates fed less voraciously, leaving more food for the focal colonies (this seems unlikely given these neighbors were larger overall, and so have higher total resource consumption rates); or (2) slower metabolism and, therefore, larger neighboring colonies, may have altered local flow regimes to benefit focal colonies more. Previous studies in this system indicate that conspecific size is a key mediator of the delivery of resources to interacting individuals Cameron et al., 2017), and we suspect size, rather than per capita resource consumption drives our results here. In aquatic systems (including our own), the physical structure of sessile organisms can disrupt boundary currents and increase resource entrainment, particularly when water currents are too fast Okamura, 1984;Svanfeldt et al., 2017).
Thus, it is possible that focal colonies likely benefited from being adjacent to fast growing, low-metabolic rate neighbor colonies on panels where flow was higher as they baffled the current more. We also found that the effects of the metabolic rate of neighbors differed in persistence and sign across our replicated panels-we suspect this variable effect arises because of small scale differences in currents.
On higher flow panels, focal colonies may have benefited from low metabolism, large neighbors baffling flows, but on low flow panels, focal colonies suffered in the presence of such baffling (Svanfeldt et al., 2017).
We only investigated pairwise interactions between conspecifics, but intraspecific interactions occur across a range of densities in nature. Population density has been shown to affect the mode of competition (Cameron et al., 2007) as well as mediate transitions between competition and facilitation among species at least . Similarly, the frequency of a given metabolic phenotype within a population may alter the outcome of interactions among conspecifics (Ayala & Campbell, 1974). Therefore, an important next step would be to orthogonally manipulate both the density and frequency of individuals of known metabolic phenotypes within a population and test for facilitative and competitive interactions.
Among species, context-dependent changes in the strength of competitive interactions are an important maintainer of species coexistence (Chesson, 2000a(Chesson, , 2000bHart & Marshall, 2013).
Similarly, it would be reasonable to expect that spatial variability in strength and direction of conspecific interactions maintains within-population variation in metabolic rate (Pettersen et al., 2020). Although we found differences in conspecific interactions among microenvironments, the metabolic phenotype of focal colonies covaried with growth but had little effects on their survival or reproductive fitness. Specifically, we found that the focal metabolic rate affected the survival of focal colonies on one panel (interacting with the metabolic rate of the neighboring colony), but we could not detect an effect of the focal metabolic rate on either survival or reproductive outputs on the other panels.
Instead, the performance of focal individuals on these panels was more consistently affected by the metabolic phenotype of their neighbor. Neighbor metabolic rate affected the survival, growth, and reproduction of focal colonies, albeit in contrasting ways that varied among microenvironments. These results suggest that complex eco-evolutionary feedbacks (akin to "indirect genetic effects," sensu Wolf et al., 1998) are likely to maintain variation in metabolic rate despite previous studies showing strong directional selection (that should erode phenotypic variation) on focal metabolic rate in this system (Pettersen et al., 2020). An important next step will be to determine the degree to which metabolic rate is heritable in this system.
That the metabolic rate of the neighbor colony had more pervasive effects on focal colonies than did the metabolic rate of those colonies themselves surprised us. Most studies to date have focused on covariance between the focal organism's metabolism and the performance of that organism (Pettersen et al., 2018). We can find few examples of studies that explore how the metabolic rate of one individual affects the performance of other individuals (Auer et al., 2020). Yet it is well understood that metabolic rate covaries with any number of traits that determine how an organism will interact with and affect its environment and other species (e.g., body size, foraging rate, resource use; Biro & Stamps, 2010;Careau et al., 2008). Thus, in this context, perhaps our results are less surprising than they first appear.
We recommend that future studies of the ecological effects of metabolism expand their scope to investigate competitive interactions both within and among species and where possible, be done under field conditions. We predict that variation in metabolic rate is likely to have effects that extend beyond the focal organism but for the most part, these effects are unexplored. Our study highlights the importance of replicating arrays of competitors under natural conditions. We replicated our response surface design in space and found very different effects from one microsite to another-in the absence of such replication, we would have overestimated the consistency of metabolic effects and drawn potentially misleading conclusions about how metabolism affects competitive interactions. Instead, we found that, while the metabolic phenotypes of both focal individuals and their neighbors matter, their effects can differ in strength and direction-capturing this variability is necessary for a more complete understanding of such effects in nature. Future studies are necessary to determine why we see such variable effects of metabolic rate among microenvironments, but we suspect a small-scale variation in current regimes and the delivery of resources-future studies will manipulate local food availability to determine its role (Svensson & Marshall, 2015).

ACK N OWLED G M ENTS
We thank the Royal Melbourne Yacht Squadron and the Royal Brighton Yacht Club for access to our field sites, Belinda Comerford for field and technical assistance, and the Australian Research Council for funding. We also thank Paul Gribben and Peter Edmunds for helpful comments on earlier versions of the manuscript.